in two-way avoidance learning

Transcription

1 Intertrial responses of rats in two-way avoidance learning to visual and auditory stimuli Kazimierz Zielin ski, Tomasz Werka and Eugeniusz Nikolaev Department of Neurophysiology, Nencki Institute of Experimental Biology, 3 Pasteur St., Warsaw, Poland Abstract. Learning and performance of two-way avoidance were investigated in a total of 68 rats trained with either a visual (change in illumination to darkness or to light) or an auditory (white noise of 70 or 60 db intensity) conditioned stimulus (CS). Experiment I showed that the darkness CS produce lower avoidance performance and a much higher rate of intertrial responses than either a auditory or a compound (visual plus auditory) CS. A monotonic within-session increase of avoidance performance and a similar, but less regular increase of intertrial responses were found at the beginning of training in each group. In later sessions such trends were observed only in rats trained with a visual CS. Experiments I1 and 111 showed also rapid transfer of avoidance response and a corresponding change of intertrial response rate with the change of CS modality. When the compound CS was used, the effects of the visual element were completely overcome by the auditory one. Rats trained with a visual CS in Experiment 1V showed a positive correlation between avoidance performance and the number of intertrial responses, which was more pronounced in earlier than in later training sessions. We consider the rise of intertrial behaviour as an adaptive response to the increase of task difficulty. CSi of different modalities differ not only in relative saliency, but also in the discriminability between their onset and offset. The modality of the CS influences not only avoidance performance but also the course of learning. Key words: two-way avoidance, intertrial response, conditioned stimulus modality, transfer test, rat

2 72 K. Zielinski et al. INTRODUCTION Studies of the effects of prior exposure to shock on subsequent avoidance learning showed that the amount of movement exhibited during preliminary shock sessions may be used as a good predictor of subsequent avoidance performance (Anisman and Waller 1971,1972). However, the same papers reported the lack of similar correlation between motor activity during preliminary test sessions and later avoidance performance in nonshocked rats. Since preshock decreases movement and increases crouching responses, these response repertoire changes may be an important determinant of subsequent avoidance performance (Cicala and Ulm 1971, Anisman and Waller 1973b). Numerous studies confirmed these general statements, revealing, however, a high variability of the consequences of inescapable shocks and resulting changes in motor activity on avoidance behaviour depending on the task, required response, species and strain, and features of the apparatus used (Alleva et al. 1983). Much less attention has been paid to the relations between activity during the action of discrete conditioned stimulus (CS) and that occurring during intertrial intervals (ITIs) in the course of avoidance learning. A typical S-like learning curve expressing an initial slow, then rapid, and again slow increase of active avoidance performance should be accompanied by an initially increasing and then decreasing function for changes in frequency of intertrial responses (ITRs) (Werka 1980, Zielinski and Plewako 1980). Group data showed that the most rapid acquisition of two-way avoidance in rats was characterized by a rapid initial increase of ITRs, whereas rats that failed to reach avoidance criterion performed a smaller number of ITRs (Abuladze and Tchutchulashvili 1981). In an experiment employing eight groups of rats trained in a running wheel avoidance, in which the effects of CS termination and avoidance of shock were compared, it was shown that the number of ITR highly correlates across conditions with avoidance performance (Bolles et al. 1966). In experiments on cats trained in bar-press avoidance we noted that the rates of avoidance responses to a weak CS at the beginning were only slightly higher than the rates of ITRs (Zielinski 1979). A similar relation was found between the rates of ITRs and of responses to the newly introduced nonreinforced CS during go, no-go avoidance differentiation training. Moreover, under some conditions these values showed a significant positive correlation (Zielinski and Czarkowska 1973). In addition, the number of ltrs was strongly correlated with an increase of the number of short-latency responses performed at the beginning of reversal training to the CS that had previously been paired with shock and then denoted no shock and was no longer terminated by an instrumental response (Zielinski and Czarkowska 1974, Kowalska et al. 1975). These data suggest that the saliency of the CS may be a variable influencing interrelations between avoidance responses (CR,,,) performed in the presence of the CS and the ITRs, when a subject is under the influence of contextual cues. Numerous experiments showed a marked disparity of different stimuli in their effectiveness as CS depending on the prevailing motivation and response requirements of the task. There are two main factors responsible for such influences: the role of different stimuli in modulating response hierarchies (Anisman and Waller 1973b) and their arousing functions in evocation of the learned conditioned response (Zielinski 1979). In rats trained in discrete trial active avoidance, the superiority of noise over light CSi has been well documented not only for two-way avoidance (Thompson and van Hoesen 1967, Morris 1974, Jacobs and LoLordo 1977) but also for bar-press avoidance tasks (D'Amato et al. 1964, Myers 1964, Biderman 1967). Bignami and his group demonstrated in varieties of go, no-go avoidance differentiation tests that higher effectiveness of auditory over visual CSi in the shuttle-box went in complex interaction with response and reinforcement factors (Rosid et al. 1969, Rosid and Bignami 1970, Frontali and Bignami 1973, 1974, Bignami 1980). The aim of this paper was an examination of the auditory and visual stimuli effects on two-way avoidance in rats. In contrast to previous studies, we were interested not only in the comparisons of performance levels of the CR,,, but also in the numbers of ltrs at different stages of avoidance training. Since complex interactions between CSi and other variables play an important role in active avoidance performance (Bignami 1989), any inference based on comparisons of the results obtained by different investigators may be severely biased. Accordingly, we decided to conduct a series of experiments in which the effects of visual and auditory CSi on two-way avoidance were compared using between- and within-subject designs. In Expt I the effectiveness of visual, auditory and compound CSi were compared in three groups of rats. In Expts 11 and 111 rats were trained with two CSi in succession, visual and auditory, using a counterbalanced order of stimuli in two groups of subjects. To ease the transfer of

3 Avoidance and ITR 73 the acquired avoidance, a session was inserted between the two stages of training in which the two stimuli, the one used already and a new one, were given in a compound. Thus, in each group of rats Kamin's (1969) design for studying the blocking effect was used. Although a typical blocking effect in instrumental active avoidance training was not expected, response levels to the compound CS and during ITIs in this transfer session were of special interest. In Expt IV all rats were trained with only the single visual CS, to collect additional data on the correlation between the CR,, and ITR performance. EXPERIMENT I. TRAINING TO DARKNESS, THE 70 db AUDITORY NOISE OR THE COMPOUND: ASYMPTOTIC PERFORMANCE Method The experiment was conducted with 24 adult male Moll-Wistar rats bred in the Institute, experimentally naive and weighing g. Subjects were kept in groups of eight in a single home-cage (43 cm long, 25 cm wide, 18.5 cm high), containing food and water available ad libitum. A normal light-dark cycle was maintained, and rats were trained once a day in the morning or early afternoon, according to the same order and about the same time of the day. A shuttle-box apparatus consisted of two identical opaque dark Plexiglas compartments (31 cm long, 18 cm wide, 29 cm high) separated by a wall with a rectangular (7 cm wide, 10 cm high) opening with a sill situated on the grid-floor level. Each compartment was covered by a movable transparent Plexiglas ceiling and illuminated by a 5 W lamp mounted centrally just below the ceiling. On each wall opposite to a central partition, a 10 cm loudspeaker was mounted outside of the apparatus and 15 cm above the floor. Crossing through the opening was recorded by light photocells mounted 4 cm to either side of the central partition, 5 cm above the floor level. The floor in each compartment was constructed from 16 stainless steel bars, 0.4 cm in diameter, and located parallel to the central partition 1.5 cm apart from each other. The shuttle-box apparatus was placed in a dark, sound-proof room. Subjects' behaviour was watched on a TV monitor in an adjoining room in which equipment for automatic programming of the experiment and recording of data was located. PROCEDURE Prior to avoidance training each rat was habituated to the situational cues of the apparatus for 10 min on two consecutive days. At the beginning of each session a rat was placed in the left compartment of the shuttle-box, close to and facing the end wall. After 20 s a trial started with CS onset and, 5 s later, the 1.6 ma scrambled shock activated the grid-floor (unconditioned stimulus, US). Running to the opposite compartment within the 5 s CS-US interval precluded the foot-shock, immediately terminated the CS. and was scored as an avoidance response. Running to the opposite compartment after the US onset immediately coterminated the CS and US and was scored as an escape response. Each daily training session consisted of 50 trials. The intertrial intervals were of s duration (mean = 20 s) and varied in a semi-random order. During intertrial intervals subjects were permitted to move in any direction, so they could cross away from or back into the compartment in which they had been previously. The next trial always started in the compartment in which the subject was actually present. The latencies of instrumental responses, the numbers of avoidances and ITRs, and the durations of US action were recorded. GROUP TREATMENT Before habituation sessions rats were randomly assigned to three groups, 8 subjects each. During the nine daily training sessions, the groups differed only in the CS used; Group D was trained with darkness as the CS, the Group N was trained with the onset of a 70 db (re: 20pN/m2) wide bond (white) noise as the CS, and both those stimuli given in compound served as the CS for the Group ND. The darkness CS was provided by offset of the ceiling light in the compartment occupated by a rat whereas the opposite compartment was illuminated as before. Just after a rat left the shock compartment it was again lighted and both compartments were illuminated during intertrial intervals. In Group N both compartments were continuously illuminated the whole session. After training sessions some additional test sessions were conducted. For half of the rats from Group D and from Group N during

4 74 K. Zielinski et al. Session 10, a compound CS of darkness plus 70 db noise was used, while the remaining rats in each group were given additional training with their original CSi. All rats from Group ND were switched during Session 10 from the compound CS to the darkness CS, then during Sessions 11 and 12 trained with the compound CS, and again switched to the darkness CS during Session 13 and either returned to the compound CS or trained with the darkness CS during Session 14. Results During the habituation sessions, the most frequently observed behaviours were crossings and rearings. The overall mean number of crossings from one compartment to the other during the 10 min session was 13.2 (range 2-28 for individual rats and sessions). No marked differences between groups or days were found for the numbers of rearings (mean= 2.6, range 0-11). The numbers of crossings during the habituation days were compared with the numbers of ITRs on the first day of avoidance training. Normalization of the data for time (i.e. 10 min habituation sessions and 16.6 min of the overall intertrial interval periods during training sessions) showed that the number of crossings decreased over consecutive days. A 3 x 3 ANOVA showed a significant day effect (F21,2=12.02, P <0.001) and further Duncan tests yielded differences between habituation Sessions 1 and 2 (P<0.01) and between habituation Session 1 and training Session 1 (P<0.001). Although the main group effect was not significant, the between-days differences were pro- nounced in Group N and absent in Group D, resulting in a significant interaction of the two main effects (F4/,, = 3.28, P < 0.02). Duncan Tests for this interaction showed that in Group N the rate of crossings during habituation session 1 differed from that during habituation Session 2 (P<0.01) and the ITR rate during the subsequent session (P< 0.001). In Group ND the rate of crossings during habituation session 1 differed only from the ITR rate during the first training session (P < 0.01). The avoidance performance presented in Fig. 1 clearly reflects the superiority of rats trained with the auditory CS or the compound CS over rats trained with only the visual CS. This difference was observed from the very beginning of training and was maintained after reaching asymptote. A 3 x 9 ANOVA showed main effects of Group (F2/,, =28.86, P<0.001), and Day (F,/,,,=78.95, P<0.001) and a significant interaction (F,,/,,, = 1.87, P < 0.05) between them. Further Duncan Tests showed that the Group D differed from both other groups at the P <0.001 level, whereas Group N and Group ND were similar, and these relationships were very stable. On each consecutive day of training, Group D significantly differed from both Groups N and ND, mostly at the P<0.001 level, and in not one session were there observed any differences between the Groups N and ND. In not one group of rats did asymptotic avoidance performance reach the 100% level. This finding was related to yithin-session variability of the CR,, performance. Comparison of the numbers of CR,, in consecutive 10-trial blocks showed a monotonic increase Fig. 1. Changes of avoidance responses in Group D (thin line), Group N (heavy line) and Group ND (dotted line) in the consecutive sessions of ~ x~t'1. Fillied diamonds denote the darkness CS, circles denote the 70 db noise CS. filled diamonds in circles denote the compound CS consisting of 70 db W a I I I I I 1 I I 1 noise and darkness

5 Avoidance and ITR 75 Fig. 2. Changes of avoidance responding in five 10-trials blocks of each session of Expt I. Symbols as in Fig. 1. of avoidance performance within each session. As seen in Fig. 2 those differences were very pronounced at the beginning of training and gradually decreased. However, Group D was marked by the lowest numbers of CR,, performed in Block 1, which never reached more than 65% of avoidances attained by the Group ND in the same session and block of trials. Similar, but even more pronounced differences in the numbers of ITRs were observed. As seen from Fig. 3 the level of ITRs was much higher in Group D than in the other groups. A 3 x 9 ANOVA yielded significant Group (F2,,, =52.58, P<0.001), Day (FEI1 68 = 7.19, P < 0.001) and their interaction (F = 5.89, P <0.001) effects. The TTR level of Group D differed significantly from those of the other groups (P's <0.001). This was true for each consecutive day of training (P's < 0.05 on Session 1 and P's <0.001 for each consecutive session). Within Group D the lowest level of ITRs was observed during Session 1 which differed from the ITR levels of each subsequent session (P's < 0.001, except for Session 8, when only a P < 0.05 difference was observed). After reaching a maximum

6 76 K. Zieliliski et al. Fig. 3. Numbers of intertrial responses in the consecutive sessions of Expt I. Symbols as in Fig. 1. during Session 3, a decrease in the numbers of ITRs was observed with difference between Sessions 3 and 4 (P < 0.05), and even stronger differences found between Session 3 and Sessions 5-9 (P's<0.001). In the other groups no differences in the levels of ITRs on consecutive days were observed. Generally, the level of ITRs was lowest at the beginning and highest at the end of each session. Such a trend was very marked in Group D and only noticeable in the other groups characterized by low overall levels of ITRs. The results of the additional test sessions are presented in Table I and Fig. 4. Before test sessions rats from each group were randomly assigned to a Subgroup a, which was trained during the test session with the same CS as during the previous session, and to Subgroup b, which was trained during the test session with the compound CS. Note that for the Group ND such final test session was given after two changes from the compound CS to the darkness CS. In Fig. 4 changes in the CR,, performance are presented for each individual rat of Group ND whereas all other data are group or subgroup means. The results are very clearcut. Each switch from the compound CS to the darkness CS resulted in the decrease of the CR,, performance and a marked rise of the ITR level, and each change from the darkness CS to the compound CS produced opposite effects. Data presented in Fig. 4 indicate that the second transfer from the compound CS to the darkness CS elicited smaller, although reliable changes in the CR,, performance and the ITR level. Change from the 70 db noise CS to the compound CS was without any consequence in overt rats' behaviour. Discussion The results of Expt I show marked differences in CR,, performance and ITR level between rats trained with the darkness CS and the 70 db noise CS. Adding the visual element to the 70 db noise CS exerted no effect on behaviour, whereas adding the auditory element to the darkness CS resulted in immediate changes in the CR,, performance and the ITR level. These effects, found during the additional test sessions, which were rapid in occurrence and observed in each individual rat, may be interpreted as changes in the performance rather than differences in learning functions. Similar results were observed in classic experiments concerning stimulus intensity effects on response strength employing factorial designs (Hillman et al. 1953, Kessen 1959). Thus, it may be

7 Avoidance and ITR 77 TABLE Comparison of mean avoidance (CR,,) performance and mean number of intertrial responses (ITR,) for each subgroup during the last test session and the preceding session of Expt I. The CS was not changed in subgroups D-D, N-N and (N)D-(N)D, whereas in subgroups D-DN, N-ND and (N)D-ND the compound instead of the single CS was presented during the test. Note that before the tests the subgroups D-D and D-ND were trained during nine sessions with darkness CS and the subgroups N-N and N-ND were trained with 70 db noise CS. The subgroups (N)D-(N)D and (N)D-ND were trained with the compound CS but during the last session before the test they were trained with darkness CS Subgroups Sessions compared CR,,, TTR Subgroup D-D D-DN Subgroup N-N N-ND Subgroup (N)D-(N)D (N)D-ND Fig. 4. Changes of avoidance and intertrial responses during test sessions in Group ND of Expt I. In the upper part, the percentages of CR,, are presented for individual rats, in the lower part, the numbers of 1TRs for the entire group are given. Symbols of CSi as in Fig. 1. assumed that the observed effects resulted from changes in saliency of the triggering stimulus in overtrained subjects. However, from the very beginning of avoidance training visual and auditory stimuli exerted different effects on the free "operant7' level of crossing the central partition of the shuttle-box. The introduction of pain signalled by an auditory stimulus resulted in a decrease of the number of spontaneous crossings (now considered as ITR). In contrast, the introduction of pain ignalled by a visual stimulus immediately arrested the decrease of spontaneous crossings. In the next training sessions, the ITR level observed in rats trained with the darkness CS was markedly higher than the rate of crossing during the habituation sessions. These observations ought to be compared with numerous studies showing a depressive effects of inescapable shocks on subsequent motor activity measured in the absence of shock (Pinel et al. 1971, Anisman, Waller 1973a, Allewa et al. 1983). Signaled shocks also suppressed motor activity, however tests in shuttle-box showed that crossing responses were more frequent during the CS action than between CS presentations (Cicala et al. 1971, Cicala and Ulm 1971). Moreover, it occurred that light CS produced less freezing that the noise CS (Sigmondi et al. 1980). It should be stressed that from the very beginning of training all effects of the visual component were completely overshadowed by the auditory component presented in a compound. All of these data and considerations seem to suggest that not only performance but also the learning functions differed between groups of rats trained with the darkness or the auditory noise CSi.

8 78 K. Zieliriski et al. EXPERIMENT 11. DARKNESS VS. 70 db ACOUSTICAL NOISE: RECIPROCAL TRANSFER The aim of Expt I1 was to test the extent of transfer among the darkness CS, the acoustical CS of 70 db intensity noise, and the compound consisting of both elements. In contrast to the tests conducted at the end of the Expt I the transfer effect was investigated before the avoidance performance reached asymptote. Method TABLE II The design of Expt I1 and Expt 111 and the quality of CSi used at consecutive stages of training. Abbreviations: D, darkness; L, light; 70 N, noise of 70 db intensity; 60 N, noise of 60 db intensity Group Sessions 1-4 Session 5 Sessions Expt I1 Group D-DN-N D D+70N 70N Group N-NL)-D 70N D+70N D Expt 111 Group L-LN-N Group N-NL-L The experiment was conducted on 16 adult, experimentally naive male Moll-Wistar rats weighing g obtained from the same source and maintained identically as those used in Expt I. Apparatus and Procedure were the same as in Expt I. GROUP TREAZMENT Prior to the experiment, the rats were randomly assigned to two groups 8 subjects each. During each of two habituation sessions, subjects were placed individually for 10 min into the shuttle apparatus with the house light on, and the number of crossings through the central opening was recorded by light photocells. Avoidance training started on the next day. For Group D-DN-N offset of the house light was used initially as the CS, and during Session 5 the darkness periods were presented in compound with the 70 db noise. For sessions 6-9, the 70 db noise alone served as CS. Group N-ND-D started training with 70 db noise CS, given in compound with darkness during Session 5, and finished training with the darkness CS presented alone. Each CS was presented in the compartment occupied just prior to the trial, so that subjects moved away from the CS during training. The arrangements of CSi used in each group are shown in Table 11. Results The changes of the avoidance performance levels presented in Fig. 5 for both groups clearly differentiate the three phases of the experiment. Rats trained during sessions 1-4 with the noise CS performed more avoid- ance responses than did rats trained with darkness CS. The two groups equalized their performance during Session 5 when the compound stimulus, noise and darkness, was used for all rats. Performance during sessions 6-9 indicated that the rate of avoidance responses was determined by the quality of the CS actually used. A 2 x 9 ANOVA of the avoidance performance on each day revealed an effect of Days (FBI,, = 72.07, P < 0.001), a significant Days x Groups interaction (FBI,,, = 10.55, P<0.001) but no main effect of Groups. Further Duncan Tests showed differences in avoidance performance between the two groups in Session 4 (P < 0.05) and sessions 6-9 (all P's at least<o.ol). In each group avoidance performance during Session 1 was lower than in any following session (P< 0.001) and the same was true for session 2 and subsequent sessions (P<0.05 or better). Another measure, the total time of shock presentations in consecutive sessions, changed in parallel to the avoidance performance. A 2 x 9 ANOVA revealed an effect of Days (F8/,,, = 31 33, P <0.001), a sig- nificant Days x Groups interaction (FSl112=2.00, P<0.01), but no main effect of Groups. The Duncan Tests yielded further interesting information from this measure. During Session 1, subjects trained with the visual CS (Group D-DN-N) experienced more shock (P<0.001) then those trained with the auditory CS (Group N-ND-D). A similar, although not significant, differences related to the CS quality were observed in next sessions. During sessions 2-5, more shock occurred in Group D-DN-N; during sessions 6-9, in Group N-ND-D. The number of intertrial responses was also directly related to the modality of the CS used at different

9 Avoidance and TTR 79 stages of this experiment. As seen from Fig. 6 the level of TTR was high when rats were trained with the darkness CS and low when the noise CS was used. As with the other measures, the 2 x 9 ANOVA revealed a Days effect (F,/,,, = 4.48, P <0.001), a significant interaction (F,/,,, = 17.69, P< 0.001), but no effect of Groups. Further Duncan Tests indicated significant between group differences on Session 2 and all subsequent sessions. Consistent with the switch in CSi, the two groups differed in the number of ITRs only when sessions 1-4 were analyzed. A 2 x 4 ANOVA showed an effect of Groups (F,/, = 9.57, P <0.01) and Days (F3,,, = 4.39, P < 0.01), whereas the interaction of the two main effects was not significant. Similarly, the two groups differed also during sessions 6-9, analyzed separately (Groups: F, /, = 63.74, P <0.001; Days: F,/,, = 3.63, P <0.02; interaction: F3,,, = 3.89, p < 0.02). Fig. 5. Changes of avoidance responding in Group D-DN-N (thin line) and Group N-ND-D (heavy line) of Expt 11. Syrnbols of CSi as in Fig L Fig. 6. Numbers of intertrial responses in the consecutive sessions of Expt 11. All symbols as in Fig. 5.

10 80 K. Zieliriski et al. There were marked changes of performance in the course of each session, with the smallest number of avoidances in the first block of 10 trials and the greatest in the last block. The largest within-session variability in avoidance performance was observed during Session 3 with 17.5% of avoidances in Block 1 and 75% of avoidances in Block 5 in Group D-DN-N, and 17.5% versus 90.0% of avoidances for Group N-ND-D, respectively. Thus, the high density of the shock at the beginning of each session resulted in rapid improvement of the avoidance performadce. Only during Session 9 in Group D-DN-N was the difference between Blocks 1 and 5 as small as 10% of avoidance responses. A similar decrease of within-session variability in performance level with prolonged training was not observed in Group N-ND-D. The numbers of ITRs displayed a similar, although less regular, increasing within-session trend, but only when subjects were trained with the darkness CS. The data presented in Fig. 7 illustrate the rapid changes of the rat's behaviour during intertrial intervals relative to changes of the CS modality during the course of the experiment. The most interesting is the TTR function for Group D-DN-N during Session 5, when the compound CS was introduced. The numbers of TTRs emitted in Blocks 1 and 2 resembled the trend from the previous sessions, when rats were trained with the darkness CS. Already by Block 3 the number of ITRs decreased, and in Block 5 resembled the TTR level -- TABLE 111 Group mean percentages of avoidance responses performed in trials 1-10 (Block 1) of the Sessions 4, 5, and 6 of Expt TI and Expt TIT Group Sess~on 4 Session 5 Session Expt I1 Group D-DN-N Group N-ND-D Expt TIT Group L-LN-N Group N-NL-L typical for subjects trained with the noise CS. As far as Group N-ND-D was concerned, the most interesting finding was the highly elevated levels of TTRs from the very beginning of Session 6, when rats had to respond for the first time to the darkness CS presented alone. The changes in the avoidance performance during sessions 5 and 6 were much less evident in both groups. However, comparisons of the avoidance performance during the first blocks of sessions 4-6 presented in Table I11 indicated that the behaviour changed rapidly with the new stimulus conditions. Subjects of Group D-DN-N increased CR,, level immediately when confronted with the darkness-plus-noise compound CS. The avoidance performance level was further enhanced Fig. 7. Numbers of intertrial responses in five 10-trials blocks of each session of Expt 11. All symbols as in Fig. 5.

11 Avoidance and ITR 81 when the noise CS alone was given during Session 6. On the contrary, there were no increase of the CR,, levels between Sessions 4 and 5 in Group N-ND-D, originally trained with the noise CS. In this group the CR,, performance markedly decreased in Session 6, when the subjects were confronted for the first time with the darkness CS presented alone. Discussion The results confirm marked differences between groups of rats trained with the auditory CS or the visual CS in avoidance performance and ITR frequency. As in Expt I the behaviour of rats confronted with auditory and visual elements given in a compound stimulus was similar to that to the auditory CS alone. In accordance with expectations, no blocking effect was observed. In fact two distinct avoidance performance functions were obtained. Both the avoidance performance and ITR frequency were fully determined by the CS actually used. Rats trained originally with the visual and then switched to the auditory CS immediately reached the avoidance performance level that would be expected for.rats trained with the auditory CS all the time. The data for the first trials of the sessions presented in Table I11 supported such a conclusion, so that it may be said that performance underestimate learning, for the Group D-DN-N at least. However, one set of data suggested that some learning was involved during transfer from one CS to another. Subjects trained initially with the darkness CS reduced the number of ITRs only after 20 presentations of the compound, darkness-plus-noise CS. On the other hand, the ITRs were very numerous from the very beginning of Session 6 in subjects trained initially with the auditory CS and switched to the visual CS presented alone. At the beginning of training with the visual CS, elevated ITRs rate occurred in Group D-DN-N only toward the end of Session 1. These findings suggest that pronounced between-group differences in the number of ITRs emerged in the course of training. EXPERIMENT 111. LIGHT VS. 60 db ACOUSTICAL NOISE: RECIPROCAL TRANSFER It was planned to repeat the Expt I1 testing the effects of light CS instead of darkness CS and noise of 60 db instead of 70 db intensity. We expected that the effects of the house-light offset CS would be similar to the house-light onset CS whereas saliency of the 60 db noise will be less that the 70 db noise. Thus smaller between-group differences were expected than in Expt 11. Methods The experiment was conducted on 16 adult, experimentally naive male Moll-Wistar rats weighing g obtained from the same source and maintained identically as those used in Expts I and 11. Apparatus and Procedure were the same as in Expts I and 11. GROUP TREATMENT Before the experiment, rats were randomly assigned to two groups 8 subjects each. Avoidance training started after two habituation sessions during which each rat was placed individually for 10 min into the apparatus with the house-light turned off, and the number of crossings through the central opening was recorded by photocells. For Group L-LN-N at the beginning of training, house-light onset in shock compartment was used as the CS, during Session 5 the light was presented in compound with the 60 db noise, and during sessions 6-9 the 60 db noise alone was the CS. Group N-NL-L started training with the 60 db noise CS, given in compound with light during Session 5, and finished training with the light CS presented alone. Each of these CSi was presented in the compartment occupied by a rat just prior to the trial, and moving away from the CS was required. The CSi used in each group are presented in Table 11. Results For each rat the numbers of crossing and rearing were scored during each of the two habituation sessions. The overall mean number of rearing was 4.0 (range 0-10 for individual subjects and sessions) with no differences between groups or days. The overall mean number of crossings from one to the other compartment during the 10 min period was 13.8 (range 4-23 for individual rats and sessions). When the numbers of crossings during the habituation days were compared with the numbers of ITRs during the first

12 82 K. Zieliriski et al. Fig. 8. Changes of avoidance responding in Group L-LN-N (thin line) and Group N-NL-L (heavy line) of Expt 111. Circles denote the 60 db noise CS, open diamonds denote the light CS. day of avoidance training, it appeared that after normalization of the data for the two time periods (10 min habituation sessions and 16.6 min of the overall intertrial interval periods during training sessions) the beginning of avoidance training resulted in some decrease of the spontaneous crossing level. A 2 x 3 ANOVA showed no effect of Groups, whereas the effect of Days was on the border of significance (F,,,, = 3.08, P = 0.06). There was no interaction. Comparisons of avoidance performance (Fig. 8) and the numbers of TTRs (Fig. 9) with those for Expt I1 indicated that the results obtained in both experiments were nearly identical. One may notice a slightly higher level of avoidance performance to the light CS used in Expt I11 than to the darkness CS employed in Expt 11. Similarly, the number of TTRs was somewhat higher with the 60 db noise in Expt 111 than with the 70 db noise in Expt 11. All ANOVA's based on data collected in Expt JTT yielded similar results as those presented earlier for Expt 11. Some variations between the two experiments were found in the Duncan Tests. When performance levels for consecutive sessions were compared, the between-group differences in Expt I11 were smaller than those observed in Expt 11, except for the number of ITRs. Similar to the results of Expt 11, mean percentages of CR,, for each of the Blocks 1 of the Sessions 4-6 were estimated and presented in Table TIT. In Group L-LN-N a marked increase of avoidance performance was observed after the introduction of the light plus noise compound CS during the Session 5. Further rise of CR,,, level was observed during the next session, when only the 60 db noise CS was used. When subjects of Group N-NL-L were confronted for the first time with the light CS presented alone, a marked drop of avoidance performance was observed. Discussion The similarity of the results of Expt I11 to those of Expt 11, although expected, were nevertheless astonishing. The resemblance of the figures has to be considered as evidence of reliability of the CS effects. Moreover, the results of both experiments analyzed together show that differences between auditory and visual CSi should be considered as CS modality effects on two-way avoidance. The saliency or intensity effects of the CSi were superimposed on the basic differences related to the modality of the CSi. For instance, there were no differences in CR,, performance between rats trained with the auditory CSi in these two experiments, however, rats trained with the more intense (70 db) noise CS performed less ITRs than rats trained with the less intense (60 db) noise. The highest ITR frequency was observed in rats trained with the visual CSi. The transfer from the intense auditory CS (70 db) to the darkness CS produced a more dramatic increase in the number of ITRs than a similar transfer from the less intense (60 db) noise to the light CS.

13 Avoidance and ITR 83 Fig. 9. Numbers of intertrial responses in the consecutive sessions of Expt 111. All symbols as in Fig. 8. Similarly, comparisons of the levels of CR,, performance at the very beginning and the very end of training suggested that the light CS is more salient than the darkness CS. However, this difference was possibly related to the higher ITR frequency in early training sessions observed in subjects trained with the light CS in Expt 111 than in rats trained with the darkness CS in Expt 11. EXPERIMENT IV. CR,, PERFORMANCE AND ITR FREQUENCY: CORRELATIONS Data from Expts I1 and 111 were used for preliminary testing of whether there existed any correlation between the number of CSav and the number of ITR performed in a given training session. All Spearman rank correlation coefficients estimated for the early training sessions were positive and some of them were statistically significant. However, marked variation in coefficient magnitudes was observed, probably due to the small number of rats in each group. The aim of Expt IV was to test for the correlation between CR,, performance and ITR frequency on a larger sample of rats trained under identical conditions. Methods The experiment was conducted on 12 adult, experimentally naive male Moll-Wistar rats weighing g obtained from the same source and maintained identically as those used in other experiments reported in this paper. Six subjects were kept in each of the two home-cages. The apparatus and procedure were the same as in other experiments reported here. In this experiment there was only one habituation session lasting 10 min. Then all rats were trained in two-way avoidance with the darkness CS for eight sessions. Results Figure 10 shows changes of the avoidance performance and the ITR frequency in the course of training. Both functions are presented as the probability of response (CR,, or ITR) per second. Estimation of these probabilities were done in a following way. A subject during a given session had a total of 250 s (5 s CS-US interval and 50 trials) during which avoidance responses (no more than 50) could be performed. If a rat

14 84 K. Zieliriski et al. performed 25 avoidance responses, the probability of CR,, per s was 0.1 for this session. Similarly, the sum of TTT's during one session was 1,000 s. If a rat performed 10 ITRs, the probability of ITR per s was 0.01 for this session. In the same way the probability of spontaneous crossings during the 10 min habituation session was estimated and indicated in the Fig. 10. These functions show that the probability to perform CR,, was 10 times higher at the beginning of training and 25 times higher at the end of training than the probability to perform ITR. On any stage of training the probability of CR,, was higher, and the probability of ITR was lower, than the probability of spontaneous crossing the central partition during habituation session. It should be mentioned, however, that according to the indices presented on the right side of the graphs the ITR frequency was in this experiment markedly lower than in groups trained with visual CSi in Expts The results of the search for the relationship between CR,, and ITR scores are presented in Table TV. As seen from the left side of the Table the Spearman rank correlation coefficients between the numbers of CR,, and TTR performed during the same session were statistically significant early in the training and then the correlation decayed. However, when we transformed the TTR data summing all TTRs performed from the very beginning of training including a given reference session, the correlation of such score with the HAB Fig. 10. Changes of avoidance (upper graph) and intertrial (lower graph) responses in the consecutive sessions of Expt 1V. The single point on the left denotes the rate of spontaneous crossing during the habituation session. TABLE IV Values of the correlation coefficient between the numbers of avoidance and intertrial responses in consecutive training sessions of Expt IV. On the left side are shown the Spearman rank order correlation coefficients (r,), and probabilities of their significance, between the numbers of CR, and ITRs performed during a given training session. On the right side of the Table are shown the r, and P between the number of CR,, performed during a given session and the cumulative number of ITR, performed during the all previous and a given reference training session Sessions rs P Sessions 's P CR,, ITR C%, ITR

15 Avoidance and TTR 85 number of CR,, performed during the reference session was significant even at the end of training. Discussion Both methods used for estimation of the degree of association between CR,, and TTR scores indicated that at the beginning of training, the numbers of ITR and CR,, were positively correlated. The results obtained with the second method, in which cumulative numbers of ITR were correlated with the number of CR,,, performed in a given reference session, provided convincing evidence that TTR performed early in training exerted a strong effect on avoidance performance in later training sessions. The method of summing of the numbers of TTR employed here may be considere d as a transformation of an initially increasing and then decreasing function for ITRs in the course of learning into a S-like function similar to the typical learning curve of CR performance. Tn this experiment an initial increase and then a slow decrease of ITRs were observed similar to that found in rats trained with the darkness CS of Expt I. However, the overall ITR frequency in Expt IV was markedly lower than in the other groups of rats trained in this study with visual CSi. Due to this abnormality the samples of the data used for detection of the correlation between ITR and CR,,, performance may be considered to some degree as representing not only rats trained with visual but also with auditory CSi. Preliminary estimations of correlations based on scores obtained in Expts I1 and TI1 are in agreement with such a notion. GENERAL DISCUSSION Our data provide convincing evidence of the nonequivalence of auditory and visual CSi for rats trained in two-way avoidance. Two distinct learning functions were found depending on the modality of stimuli. Auditory CSi resulted in high avoidance performance and low TTR frequency, whereas visual CSi resulted in low avoidance performance and very high frequency of TTRs. Expts TT and TTI revealed that the exchange of the CSi resulted in immediate conversion of one learning function to another. No blocking effect was observed, either when more salient auditory, or when less salient visual element, were added to the CSi used in original learning. Easy bidirectional transfer suggests that both avoidance and intertrial responses were controlled entirely by the CS actually used. The results are in full agreement with those of other investigators, who used a variety of experimental design s to prove the nonequvalence of auditory and visual CSi for shuttle-box avoidance. Probably the most dramatic differences in the behavioral effects of auditory and visual stimuli were noticed by Morris (1974), who used for all of his rats a tone as the CS (warning signal in Sidman-type avoidance) and a light as the feedback stimulus, since in pilot work two-way avoidance training to a light CS proved to be impossible. Such a discrepancy may be expected, since the effective ness of a light CS in two-way avoidance training depends, in contrast to an auditory CS, on many experimental variables. Long ITI's, short maximal shock duration, no partition between the two compartments and presentation of the CS only in the start compartment, all make avoidance training to the light CS easier (Bignami et al. 1985, Bignami 1989). Only the last condition applied to our experiments. Traditionally, stimulus nonequvalence has been investigated using an experimental design for overshadowing paradigm: training to the compound CS and determining response strength to each element. A design for blocking effect seems to be more fruitful. In our experiments, the addition of the visual element to the auditory CS did not produce any change in learning functions. The addition of the noise element to the visual CS immediately shortened response latencies and elevated the avoidance response level. These data are consistent with the notion that defensive CSi possess two distinct functions: they signal a dangerous situation and they energize neural processes leading to the execution of adaptive behaviours. Stimulus generalization studies have shown that this arousing function may determine performance, since stimuli more intense elicited faster two-way avoidance responses than the CS used in original training (Rohrbaugh et al. 1971). Our data suggest that signalling properties of CSi of different modalities used in these experiments were similar, but auditory and visual stimuli differed substantially in their arousing functions. Experiments on rats trained in bar-press avoidance showed that a change in illumination was even more effective than a noise CS when paired with shocks having brief on- and long offperiods (D'Amato et al. 1964). Tn avoidance response training not only CS onset but also CS offset possess signalling functions (Bolles et al. 1970). CS onset constitutes a warning signal that, if required movement is not performed after predeter-

16 86 K. Zielinski et al. mined time period, the aversive stimulus follows. The offset of the CS denotes the beginning of a safe period, the ITI, during which no nociceptive stimulation will be given. A signal function similar to the CS offset has been postulated for the proprioceptive feedback of the avoidance movement (Schoenfeld 1950). However, in the case of the locomotion reaction done back and forth in a shuttle-box, the role of such information must be seriously attenuated. In some experiments, an additional feedback stimulus was introduced starting just after the avoidance response for a period of time shorter than the duration of IT1 (Jacobs and LoLordo, 1977). While auditory stimuli were much better warning signals than visual stimuli, visual stimuli were good feedback signals. Both light and darkness as feedback signals overshadowed the noise onset but not noise offset used as feedback signals (Jacobs and LoLordo, 1977). These results indicate that the onset of the noise was a salient warning signal and its offset was also a good feedback stimulus for rats trained with the auditory CS, whereas the change in illumination was a much less salient warning signal, and the opposite change of illumination was also a weaker feedback stimulus for rats trained with the visual CS. Moreover, a negative transfer between the onset and the offset of an auditory stimulus or between opposite changes in auditory intensity have been observed in different species and experimental paradigms (Champion 1962, Zielinski 1965, 1971), whereas opposite changes in illumination of a milk glass disc showed nearly complete transfer (Logan and Wagner 1962). It would be premature to maintain that the above considerations based on studies of nonequvalence of feedback stimuli and on transfer tests provide a satisfactory explanation of the slow avoidance learning to visual CSi. However, the low asymptotic level of two-way avoidance and high frequency of ITR observed in rats trained to visual CSi, are consistent with such a notion. The high frequency of ITRs in rats trained with a visual CS may be due to inefficient discrimination of the opposite changes in illumination level signalling respectively danger and safe periods in the shuttle-box. Monotonically decreasing ITR frequency following bar press avoidance responses, along with a markedly higher ITR frequency in cats trained in conditions precluding avoidance responses immediately terminating a warning signal (Zielinski and Plewako 1980), indicated that residual fear must be considered as the main source of intertrial responses in the avoidance situation. Monotonically decreasing ITRs frequency has been reported also for rats trained in a running wheel to avoid shock (Bolles et al. 1966). The findings of no initial differences between rats trained in the shuttle-box with auditory or visual stimuli, the gradual within session increase of ITR frequency in subjects trained with visual stimuli, and the lack of such gradients in rats trained with auditory stimuli are all in agreement with this explanation of the between-group differences in ITR frequency observed presently. Some decrease of ITR frequency observed in the group trained with the darkness CS in Expt. I and similar changes in final sessions of Expts I1 and I11 indicate that the ability to discriminate opposite changes in illumination of the shuttle-box is improved with prolonged training. On the other hand, the immediate increase in ITR frequency after the change from the auditory to the visual CS resembles the dramatic rise of ITRs when the signal values of CSi were reversed after mastering an avoidance differentiation task (Zielinski and Czarkowska 1974, Kowalska et al. 1975). Such an explanation of the between-group differences in the ITR frequency, reflecting differences in the signal values of the auditory and visual stimuli in avoidance learning, seems to be contradictory to the results of Expt IV and some other data showing that the number of ITRs was correlated with avoidance performance (Bolles et al. 1966, Abuladze and Tchutchushvili 1981). Why were ITRs resulting from insufficient discrimination between danger and safe periods positively correlated with the avoidance performance level? The answer to this question seems to be quite simple. ITRs, resulting from residual fear and gradually decreasing during the IT1 period, may be considered as an index of the fear level before termination of a trial. Thus, when comparing subjects trained under identical conditions, differences in ITR numbers reflect a disparity in behavioural arousal. It has been shown that each change of the task requirements during reward training elicited an immediate rise of ITRs, and soon after solving the task the ITRs reduce in number, denoting decrease of behavioural arousal (Zielinski 1980). We presume that the rise of ITRs is an adaptive response to the increase of task difficulty. All of our rats trained with visual CSi showed a marked increase of ITRs from the first to the next daily sessions (see Figs. 3, 6, 7, 9 and 10). The positive correlation between the ITR number and the level of avoidance performance observed in early training sessions illustrates an adaptive role of behavioural arousal for solving a difficult

17 Avoidance and TTR 87 task. It was found that after lesions of the amygdalar nucleus centralis, slow acquisition of escape from shock was accompanied by a marked delay in the rise of ITR frequency (Werka 1980, Zielinski 1990). Thus, lowering of the subject's capabilities and attenuation of behavioral arousal after lesion of this limbic structure were manifested both by the performance level of the learned instrumental response and the ITR frequency. Our data indicate that avoidance performance increases within each training session. The level of avoidances at the beginning of each successive session was lower than that attained toward the end of the previous session (see Fig. 2). Insufficient retention of the avoidance response was observed at the beginning of training in all groups, but was evident later on only in rats trained with the visual CSi. The parallel between the within-session increase of avoidance performance and the ITR frequency observed in groups trained with visual CSi may be considered as further support of the adaptive role of the increase of behavioral arousal when a subject is confronted with a difficult task. The change from the visual to the compound CS immediately improved avoidance performance and, after some delay, decrease the ITR frequency (see Fig. 7). Such a delay, in the TTR frequency change, both at the very beginning of training and then during the transitory session, seems to indicate that the high ITR frequency ought to be considered as a learned adaptive response to the task difficulty. Summing up, we conclude that within- and betweensession changes in avoidance performance and the ITR frequency indicate a continuing learning process and active evaluation of the stimulus conditions by subjects trained in two-way avoidance. Such an approach to learning and avoidance performance in the shuttle box may have the potential value for other studies. Recently it have been shown that introduction of a compound, auditory and visual, CS in rats that had attained asymptotic two-way avoidance performance with visual CS, produced a marked elevation of the c-fos protoonco gene expression (Kaczmarek and Nikolaev 1990). This findings suggests an involvement of the learning process and not just a change in avoidance performance related to a new CS quality. ACKNOWLEDGEMENT This investigation was supported by Project CPBP of the Polish Academy of Sciences. REFERENCES Abuladze G.V., Tchutchulashvili N.A. (1981) Sierdtsie, aktivnoie izbieganiie i emotsii. Metsniereba, Tbilisi, p.79. Alleva E., de Acetis L., Amorico L., Bignami G. (1983) Amphetamine, conditioned stimulus, and nondebilitating preshock effects on activity and avoidance: further evidence for interactions between associative and nonassociative changes. Behav. and Neural Biol. 39: Anisman H., Waller T.G. (1971) Effects of inescapable shock upon subsequent one-way avoidance learning in two strains of rats. Psychon. Sci. 24: Anisman H., Waller T.G. (1972) Facilitative and disruptive effects of prior exposure to shock on subsequent avoidance performance. J. Comp. Physiol. Psychol. 78: Anisman H., Waller T.G. (19734 Footshock-produced excitation and inhibition of activity in rats. Anim. Learn. Behav. 1: Anisman H., Waller T.G. (1973b) Effects of inescapable shock on subsequent avoidance performance: Role of response repertoire changes. Behav. Biol. 9: Biderman G.B. (1967) Discriminated avoidance conditioning: Stimulus function in shaping and training. Psychon. Sci. 9: Bignami G. (1980) Pharmacological evidence on the specjalization of CNS mechanisms responsible for motor act inhibition by aversive events. In: The Warsaw Colloquium on instrumental conditioning and brain research (Eds. B. ~ernicki and K. Zieliriski). PWN, Warszawa, p Bignami G. (1989) To go or not to go in aversive paradigms: Preparadness and other questions. In: Aversion, avoidance, and anxiety. Perspectives on aversively motivated behavior (Eds. T. Archer and L.-G. Nilsson). Hillsdale Erlbaum, New York, p Bignami G., Alleva E., Amorico L., De Acetis L., Giardini V. (1985) Bidirectional avoidance by mice as a function of CS, US, and apparatus variables. Anim. Learn. Behav. 13: Bolles R.C., Hargrave G.E., Grossen N.E. (1970) Avoidance learning as a function of CS quality and CS termination on escape trials. Psychol. Reports 26: Bolles R,C., Stokes L.W., Younger M.S. (1966) Does CS termination reinforce avoidance behavior'! J. Comp. Physiol. Psychol. 62: Champion R.A. (1962) Stimulus-intensity effects in response evocation. Psychol. Rev. 69: Cicala G.A., Masterson F.A., Kubitsky G. (1971) Role of initial response rate in avoidance learning in rats. J. Comp. Physiol. Psychol. 75: Cicala G.A., Ulm R.R. (1971) The effects of prefear conditioning shock intensity on initial shuttle response rate. Psychon. Sci. 23: D'Amato M.R., Keller D., DiCara L. (1964) Facilitation of discriminated avoidance learning by discontinuous shock. J. Comp. Physiol. Psychol. 58: Frontali M., Bignami G. (1973) Go - no go avoidance discriminations in rats with simple "go" and compound "no go" signals: stimulus modality and stimulus intensity. Anim. Learn. Behav. 1: Frontali M., Bignami G. (1974) Stimulus nonequivalences in goino-go avoidance discriminations: sensory, drive, and response factors. Anim. Learn. Behav. 2: Hillman R., Hunter W.S., Kimble G.A. (1953) The effect of drive level on the maze performance of the white rat. J. Comp. Physiol. Psychol. 46: Jacobs W.J., LoLordo W.M. (1977) The sensory basis of avoidance responding in the rat. Relative dominance of auditory or visual warning signals and safety signals. Learn. Motiv. 8: Jakubowska E., Zieliriski K. (1979) Avoidance acquisition in cats as a function of temporal and intensity factors. Acta Neurobiol. Exp. 39: Kaczmarek L., Nikolaev E. (1990) C-fos protooncogene and neuronal plasticity. Acta Neurobiol. Exp. 50: